Improved image change detection using fractional fourier transform

Abstract

Image change detection is increasingly becoming one of the major areas of research in Image processing. The reason for this is the proliferation of images all over the globe. New technologies are largely dependent on visual data, so it has become very important to find out sharp and faster image processing techniques which also have precision. Images are significant to us because they can be an extraordinarily effective medium for the storage and communication of information. A photograph/image allows us to avoid the need for a lengthy, tedious and, ambiguous verbal narration regarding what is there in an image. The field of image change detection attempts to replicate what our brains do, for instance, the comparative analysis of two visual landscapes, in the external world to enhance our ability of visual data processing. Change Detection techniques play a very significant role in modern infrastructure. Although there are a number of methods, their applicability is restrained by limitation of the information they are evaluated upon, the type of image acquisition available, need of information to be retrieved after change detection etc. The present dissertation undertakes a study of image change detection using Fractional Fourier transforms. DFrFT technique has been used for ordinary image change detection, primarily because the advantages this technique has over other techniques. The use of DFrFT gives us an additional parameter ‘a’ which is a fractional parameter. The parameter allows more flexibility in obtaining the output. In the end, gradient co-relation has been used for classifying the changes obtained from two images depending upon the value of correlation coefficient. Two more techniques have been used to obtain image change detection. The results from these are then compared with the outputs given to us by the DFrFT technique. Parameters like recall and precision are used to judge the level of change detection. While using an image set, we obtained values of 0.43 and 0.67 with JIH, 0.56 and 0.67 with DCT and with DFrFT at parameter ‘a’ value=0.9 we obtained 0.62 and 0.87. Overall it was observed that results improved by about 30-80%. Furthermore, the research work applies DFrFT to analyse changes in the images taken via satellites. The satellite images have been analysed because of the vital role these play in the management of natural resources and maintaining the human-naturebalance. Here the results are judged using precision, recall and F score parameter values. While taking example of an image set (3), the results have revealed that approach of Fractional Fourier transform is better than the approach followed in [35] because it provided F score value of 0.92 for fractional parameter ‘a’ = 0.98 as compared to value of 0.90 with approach followed in [35].